Confessions and Exclusion of Evidence
A defendant tells the police everything during a late-night interview, no solicitor present, having been kept awake for eighteen hours. At trial, the confession is the prosecution's whole case. Is it evidence, or is it noise the jury should never hear? English law answers this question through two overlapping filters — one mandatory, one discretionary — and a separate, later-arriving doctrine about what a court may infer when a suspect says nothing at all. Together they form one of the most heavily tested, and most misunderstood, corners of SQE1 criminal practice: a solicitor who cannot distinguish "the confession must go" from "the confession might go" will misadvise a client at the worst possible moment.
Start with the basic rule. Under PACE 1984, s76(1), a confession made by an accused person is admissible against them so far as it is relevant to any matter in issue — unless the court excludes it under s76 itself. Admissibility is therefore the default; exclusion is the exception the defence must trigger.
Definition — "confession." PACE s82(1) defines a confession broadly: any statement wholly or partly adverse to the person who made it, whether made to a person in authority or not, and whether in words or otherwise (a nod, a gesture, silence in the face of an accusation that amounts to acceptance). The definition is deliberately wide — it catches partial admissions, not just full guilty statements.
Section 76(2) then gives the defence two separate levers, and this is the detail examinees most often blur: once either is raised, exclusion is mandatory, not discretionary. The judge has no balancing exercise to perform. If the prosecution cannot discharge its burden, the confession is out, full stop.
- s76(2)(a) — Oppression. The court must exclude the confession if it was, or may have been, obtained by oppression of the person who made it.
- s76(2)(b) — Unreliability. The court must exclude the confession if it was, or may have been, obtained in consequence of anything said or done which was likely to render it unreliable.
Burden and standard. Once the defence has represented to the court that a confession was or may have been obtained by oppression or in circumstances rendering it unreliable, the prosecution bears the burden of proving, beyond reasonable doubt, that it was not. This is the criminal standard applied to a preliminary, evidential question — a genuinely unusual and easy-to-forget feature of s76.
Oppression: R v Fulling
What counts as "oppression"? The Court of Appeal answered this in R v Fulling [1987] QB 426. The facts were stark: the defendant confessed to obtaining property by deception only after police told her, cell-side, that her lover had been having a three-year affair with the woman in the next cell. The Court of Appeal held that "oppression" under s76(2)(a) takes its ordinary dictionary meaning — exercising authority or power in a burdensome, harsh, or wrongful manner. Crucially, Fulling establishes that oppression requires some impropriety by the police beyond the inherent oppressiveness of custody itself. Being arrested, questioned, and held is unpleasant by design; that baseline discomfort is not oppression. What tips the scale is wrongful conduct — deception designed to break the suspect down, threats, or worse, layered on top of ordinary custody. (On Fulling's own facts, the tactic — however manipulative — did not meet the threshold, though the case is now remembered for its definition of oppression rather than its outcome.)
Unreliability: an objective test
Section 76(2)(b) asks a different question, and it is easy to run the two limbs together under exam pressure. The unreliability test asks, objectively, whether what was said or done was likely to render any resulting confession unreliable — regardless of whether the confession is, in fact, true. This is the point that trips up students who reason "but he really did it, so the confession can't be excluded." That reasoning is precisely backwards: s76(2)(b) is not interested in truth, only in the reliability of the process that produced the statement. A textbook trigger for this limb is a breach of PACE Code C — denying a suspect access to legal advice, failing to administer the caution properly, or interviewing without required safeguards. Such breaches do not automatically exclude a confession, but they are the classic evidential foundation for arguing that what was said or done was likely to produce an unreliable account.
Suppose oppression cannot be shown and the confession was not, technically, "likely to be unreliable" — but something still went wrong in how it was obtained. This is where s78(1) does its work. It gives the court a discretion to refuse prosecution evidence, including a confession, where admitting it would have such an adverse effect on the fairness of proceedings that it ought not to be admitted. The court must have regard to all the circumstances, including the circumstances in which the evidence was obtained.
| s76(2) exclusion | s78(1) exclusion | |
|---|---|---|
| Nature | Mandatory once raised | Discretionary |
| Trigger | Oppression or unreliability | Any adverse effect on trial fairness |
| Burden | Prosecution, beyond reasonable doubt | Court's own assessment of "all the circumstances" |
| Typical fact pattern | Violence, threats, Code C breach causing unreliability | Significant and substantial Code C breach (e.g., denial of solicitor access) even without proven unreliability |
A significant and substantial breach of PACE Code C — the paradigm example being denial of access to a solicitor — can found a successful s78 application even where s76 exclusion fails on the facts. This is the safety net for cases where the confession is probably reliable and there is no oppression, but the police still cut a corner serious enough to taint the trial's fairness. Note also that s78(2) preserves any existing common law exclusionary rule in addition to the statutory discretion — s78 supplements the common law, it does not replace it.
Both routes are argued and resolved at a voir dire — a trial-within-a-trial, held in the jury's absence. The judge decides admissibility as a question of law; if the confession survives and is admitted, the jury then decides, as a question of fact, what weight to give it. This division of labour matters practically: a solicitor arguing exclusion is arguing to the judge alone, on legal grounds, before the jury ever hears a word of the disputed material.
R v Mushtaq [2005] UKHL 25. The House of Lords sharpened the line between these two roles. Once the possibility of oppression or other improper means has been raised (even if the judge ultimately admits the confession), the jury must be directed to disregard it entirely unless sure it was not so obtained — even if the jury believes the confession to be true. Before Mushtaq, juries had sometimes been left simply to weigh a possibly-oppressed confession as they saw fit; Mushtaq closed that gap, aligning the jury direction with the same "sure" standard that governs the judge's own s76 gatekeeping.
Vulnerable suspects
PACE Code C requires an appropriate adult to be present during the interview of a juvenile or otherwise vulnerable suspect. Failure to provide one is a serious procedural failure that can, on its own, found an application to exclude any resulting confession — typically argued under s78, sometimes bolstering an unreliability argument under s76(2)(b) as well.
What survives exclusion
Exclusion of the confession itself does not necessarily exclude everything downstream of it. PACE s76(4) permits admission of facts discovered as a result of an excluded confession — the standard illustration is where a defendant's (excluded) confession leads police to a hidden weapon; the weapon's discovery can still go before the jury even though the confession that led to it cannot.
A separate wrinkle arises when it is not the prosecution but a co-accused who wants to rely on another defendant's confession — for example, where D1's out-of-court admission helps exculpate D2. PACE s76A allows this, subject to the same oppression and unreliability grounds for exclusion that apply under s76. But the standard of proof flips: under s76A(2), the co-accused seeking to rely on the confession need only prove it was not obtained by oppression or unreliability on the balance of probabilities — not beyond reasonable doubt. This asymmetry makes sense once you see who carries the risk: the prosecution's higher burden protects the confessing defendant from the state; a co-accused's lower burden reflects that they are a private party trying to use evidence for their own defence, not deploying the machinery of prosecution.
The right to silence is not, in England and Wales, cost-free. Criminal Justice and Public Order Act 1994, ss34–38 permits a court or jury to draw adverse inferences from a defendant's silence in defined circumstances. This is conceptually distinct from confession law — it is not about what was said being excluded, but about what was not said being held against the defendant — yet the two areas are tested together because both turn on the fairness of what happened in the police station.
The caution, unpacked. The standard PACE Code C caution — "you do not have to say anything, but it may harm your defence if you do not mention when questioned something which you later rely on in court" — is the plain-English version of s34. It warns the suspect that silence now can cost them credibility later.
Section 34: failure to mention facts later relied on
Section 34 permits an adverse inference where a defendant fails, when questioned under caution or on charge, to mention a fact later relied on in their defence at trial. But the inference is not automatic: it may only be drawn if the unmentioned fact is one the defendant could reasonably have been expected to mention in the circumstances existing at the time. That "circumstances" test pulls in a wide range of individualised factors — the suspect's age, experience, mental capacity, state of health, sobriety, tiredness, and any legal advice received.
R v Argent [1997] 2 Cr App R 27 confirmed that having received legal advice to stay silent does not, by itself, prevent an adverse inference being drawn. Legal advice is one circumstance among many for the tribunal of fact to weigh — it is not a shield. R v Beckles [2004] EWCA Crim 2766 then supplied the analytical structure for cases turning on legal advice, setting out a two-stage test:
- Did the defendant genuinely rely on the solicitor's advice to remain silent?
- Was that reliance reasonable in the circumstances?
A defendant can satisfy stage one (genuine reliance) and still fail stage two if the advice was simply convenient rather than a sound reason for silence — which is why "my solicitor told me not to answer" is the start of the s34 argument for a defence advocate, never the end of it.
Section 35: failure to give evidence at trial
Section 35 allows the court or jury to draw such inferences as appear proper from the defendant's failure to give evidence at trial, or refusal without good cause to answer a question once in the witness box. Two safeguards precede any such inference: the court must be satisfied that the defendant is aware both that the stage has been reached at which evidence could be given for the defence, and of the effect of remaining silent; and no inference may be drawn at all if the court considers the defendant's physical or mental condition makes it undesirable for them to give evidence.
Sections 36 and 37: failure to account
Section 36 permits an adverse inference where a suspect fails or refuses to account for an object, substance, or mark found on their person, clothing, or possessions, or at the place of arrest. Section 37 does the analogous work for a suspect's presence at a place at or about the time the offence is alleged to have occurred. Both sections share a procedural precondition that examiners like to test: before either inference is available, the suspect must first be given a "special warning" — an explanation, in terms the suspect can understand, of the consequence of failing to give an explanation. Skip the special warning, and the s36/s37 inference is not available, however suspicious the silence looks.
The safety net: Section 38(3)
No matter how strong the inference, it can never do the whole job of proving guilt. CJPOA 1994 s38(3) provides that a defendant shall not be committed for trial, found to have a case to answer, or convicted solely on an inference drawn under s34, s35, s36, or s37. Practically, this means an adverse inference is always corroborative, never foundational — it can tip the balance where other evidence already points to guilt, but it cannot manufacture a case out of silence alone. For the SQE1 candidate, s38(3) is the fact that resolves almost every "can he be convicted just because he stayed silent?" scenario: the answer is always no, because something else in the evidential picture has to be doing independent work.
A confession problem on the exam will typically hand you a fact pattern and ask which route excludes the evidence, if any. The disciplined approach: check s76(2)(a) for oppression (was there impropriety beyond ordinary custody, per Fulling?); check s76(2)(b) for unreliability (was there a Code C breach or other pressure that objectively risked producing an unreliable statement, regardless of truth?); if neither mandatory ground is made out, fall back to s78 and ask whether admission would be so unfair, given all the circumstances, that discretion should be exercised. Remember the burden sits with the prosecution at the criminal standard under s76, that co-accused confessions under s76A shift that burden to the balance of probabilities, that the voir dire is where all of this is fought out before the judge alone, and that even a properly-admitted confession only goes to the jury for weight, subject to the Mushtaq direction. Layer the silence provisions on top as a separate but related battlefield, and remember the one rule that resolves nearly every silence hypothetical: s38(3) means silence alone never convicts.